Reply to the letter number 78.

665601 09.09.99


           Hello, Nemo!


   Very pleased with my RAM-drive, but such intervention in 
DOS, I think, to a fall. In addition, fix a

Pentagon _NE_RABOTAET_ on kay1024. This,
of coz, to you, but, in my opinion, if so
do NoWait, then int'y can fit ... Moreover, int, in general, 
swimming God knows where - none of dem, in which

on Scorp'e border is where they should
(AGONY, Satisfaction, etc ...) not working
correctly KAY. In this regard, I
rhetorical question: _NEUZHELI_ so hard
do the normal sync? Except
addition, there is another problem: I do not _nigde_
found descriptions of the top memory
(> 256k). I am certainly not difficult to look
the scheme, but still ... There are after all
programmers, who in iron -0. Except
order: Centronix on AY - this, of course,
simple and convenient. But here I have not yet looked at the 
scheme and subsequently failed to deliver AY, could not 
understand why it does not print: (


   Therefore, I ask my letter (mark
included) to write "sort through" about:

 *) Signals (?) Q8, AUTO, a18 (at
# 7FFD), a17 (at # 1FFD), ROMS, C1
 *) The management of upper memory (it is desirable to include 
in the documentation on board if you want to support 
programmers 1024KB)  *) The appointment of contacts on the plug

RGB connector


   This is, perhaps, everything. In conclusion, I
Still thank you for the wonderful
computer and support in our difficult zx
time.



665601 09.09.99

        Dear ............. !


   Apologies for the delay in answering. Questions witnesses in 
your letter, demand for I think - large and lengthy comments. 
So: 

1. "... Intervention in DOS, I think, to
a fall. "Firstly, the body of the DOS
- 16Kbaytnaya page ROM - available space - and it can be used,
- And that was done. Secondly, the differences
in the available versions of Firmware TRDOS'a
differ from each other no less essential than TRDOS-ROM 
firmware KAY-1024 of each of them. Third, docking with OS'yu

(And hard drive, respectively) carried out in such a way 
(through RAM-disk) are the least affected, in fact, the body 
itself TRDOS'a. (If we compare with MOA-TRDOS'om, then from the 
actual TRDOS'a there is a stone was left.) In

Fourth, most importantly, this method
capacity does not contradict any ideology
TRDOS'a, no ideology ISDOS'a. This means that the system can be 
used together - complementing and enhancing each other's 
effectiveness. 

2. Mode NOWAIT/3.5MGts. When implemented
This regime is not aimed to make
COMPATIBILITY WITH PENTAGON'om! Such a misconception widely 
held among users, due to too superficial comparison of 
architectures. It is believed (and it's stupid!) That if

is WAIT, - it Scorp (or KAY, - or
Is there any other car), unless there
- It PENTAGON. Everything else, according to users, 
unimportant. Well, some more, like you, for example, say

of INT. In fact, there is,
at least three significant differences in
Architecture PENTAGON / KAY. (More precisely, in the 
microarchitecture - because basically, it concerns a subsystem 
timing and features of the video processor.) It means that even 
if we make the INT in PENTAGON'e, then the demos (or fix'y) on 
KAY'e I still do not go. (And if they go, then

not all, and with the loss of visual effects.)
It is high time to realize that when we say "demo", we mean 
PENTAGON, when we say PENTAGON, mean "trial". This tantamount 
to approval. To were demos, you need to do PENTAGON,

One to one - with all his nonsense
and limitations. At PENTAGON'e except
dem, nothing else will do. This
machine, developed in 1991-92,
had one significant difference from all
Other Spectrum'ov. Namely - Linear
PROCESSOR. (This is different, by the way, PENTAGON, and from 
corporate Speccy.) In the language of Ellochkilyudoedochki 
users - NO WAIT. To further explain what LINEAR PROCESSOR, give 
a precise definition. Let any (ANY, ARBITRARY!) As 
distinguished two classes teams: Class A and Class B. (Command 
Classes can be selected, for example - the phase

GPU at the time of their execution,
- On location in the pages of RAM (which is
In this sense, non-linear processor Speccy)
- The type of command - etc. etc. Denote the execution time by 
T, - Then for a linear processor, the equality:


   T (A + B) = T (A) + T (B). Talking is easy,
time any team from nothing
is independent and is always the same (constant).

   This property is very important for your computer. There is 
a whole class of problems - the problem of real time. " 
(Example: the formation of an infinite sequence of periodic 
pulses with some frequency (repetition period) and controlled 
duty cycle. This is a task as you might guess, is directly 
connected with the channels communication (in particular, with 
the models without a timer and interrupt, ie, very simply, and 
therefore, very cheap). 

   DEMO Mayking is also one of the tasks
Real-time. (One of many, but
because of the existence of wretchedness PENTAGONA
other problems the public has no idea).

   It just so happened that the DEMO-making
developed specifically for PENTAGON'e because there
performs the necessary condition for the demo:
LINEARITY processors. All anything, but
for software compatibility Demoprogram Another necessary 
condition is IDENTITY (no similarity, no similarity, namely, 
identical) to the video processor and synchronization system 
(ie micro machines). And what we see in the video PENTAGON'a? 
And see We curvature. 320 lines per frame - that is

for stupidity? By telestandartu number of rows
interleaved (interlaced)
313/312 lines, but certainly not in any case not
316, or 320, or 325. Trifle, but nice: - if you want to make a 
PAL (or SECAM) encoder, then be kind enough to make him The 
curve (which is why he will not go on any other car - or go, 
but the curve, because the machine itself curve. Camcorder is, 
for example, just did not like. Ie lotion - just under the 
PENTAGON. Curves because they do not correspond TELESTANDARTU. 
You can certainly assume that this telestandart curve, and the 
video processor PENTAGON'a - live. But it is not so for two 
reasons. PENTAGON'ov on the market, the maximum tens of 
thousands, and televisions, at a minimum, tens of millions - is 
in the first place. And secondly, Spectrum was placed under the 
TV, and not vice versa. Dance to be from the stove. Neither 
developer hard'a, located in his mind and memory, not would 
make the number of lines, different from 312/313 (usually do 
312 - it's easier). If a person is not a developer, and 
businessman, he may, of course, and 320 vlepit lines. This is 
called "earn on the conjuncture "(- bird language). And that

will happen next? Yes, even the soup with a cat.

3. Now about INT'a. (Under this
about a very interesting letter: see
application below. The letter was sent
firm (c) Nemo editorial board ZX-F for
reviews (for third-party evaluation). At the end of
ends, ZX-F decided not to publish).

   Probably need more time again.
From the perspective of the firm (c) Nemo, PENTAGON'a video 
processor is not standard in Spectrum'e. Standard - this 
TELESTANDART. And the Pentagon - a deviation from telestandarta 
- and very significant. Therefore, "... do not swim knows where 
.." - It Again, the investigation decision Pentagon'a

as a standard. "... Normal sync ..." - Is that, again, as in 
PENTAGON'e? Normal "sync" (correct video processor) - a video 
processor, the corresponding telestandartu and not PENTAGON'u, 
Scorpion'y, ATM, Composit'y, Robic, Compact, Contact, Profi, 
ZX-next'y, and you never know what yet. Features video 
processors of these machines - a deviation from TELESTANDARTA. 
None of them is not standard, but approximations to

standard. The degree of deviation - this is
degree of "curvature". Including mutual.

   In INT'E important two points:
 1) binding to the phase INT'a VPU
(TV-processor scan images)
 2) the duration INT'a. More precisely three points:
 3) Timing INT'a about the processor. If the front INT'a 
(transition from 0 to 1, dx = dx) "walks" (shivers) near the 
front the last clock period (period

clock) team, the sampling error (+ / -1 clock cycle) is 
guaranteed. Regardless of how many zeros in INT'e, - at least 
five - at least twenty-five. 

   Binding INT'a to the phase of the video processor,
generally speaking, arbitrary, ie developer can choose it 
arbitrarily. In KAY'e INT "attached" to the phase of the video 
processor via a signal to the COP / (active transition from 0 
to 1, ie dx). K cycles / cycle / team processor through the 
signal * M1 / (asterisk means that the signal is displayed on 
the bus). Ie INT sampled * M1 /. (This avoids hitting the front 
(dx) INT'a in Region poll interrupt processor

than eliminates uncertainty of + / -1).

   Thus, INT "attached" to the top
frame, rather than to its end. By the beginning of the frame 
means the completion of reverse Scan rate: COP / = dx, ie 
transition COP / from 0 to 1. INT "attached" to the top

frame, ie the end of the flyback HR
sweep. Probably could be done and vice versa. But now I'm 
nothing will be gone. This is a standard for KAY'ya (Because it 
does not specify provisions telestandart INT'a). 

   Now the duration INT'a. In the model
v1.3 (95g) was applied scheme of the ZX-Revue (number and page 
unknown) before, in KAY'ah - as in other machines Spectrum'a 
used RCtsepochka. Quote scheme:


          D53: 555IR16

         6

  KC / oPE RG

         9
 * M1 / C 1913

         8 Q0 X

  +5 V EO

         January 1912

       ■ SI Q1 X

         2

       ■ 0 11

         3 Q2 X

           1 680

         April 1910

           2 Q3 * INT /

         5

           3

                     displayed on the bus (*)


   From the scheme it is clear that INT / NOT DIRECT activated 
(goes to logic 0) after passage of the COP / 1. This event 
occurs only in a cycle of M1 / (extract KOP'a) with the 
transition line / signal * M1 / CPU in the log. 0. As already 
mentioned, it eliminates the sampling error + / -1 Cycle. 
Further, in the end of the command (the last bar of the last 
cycle, namely its edge - ie, transition from CLOC'a CPU

0 to 1: * CLOCK = dx) is a survey line / signal INT / processor 
and either initiated interrupt (with the elaboration of 
confirmation on the line signal M1), or interrupt are ignored 
if they are banned. 

   Further, if we worked through the first team from the cell 
where redirected processor interrupt, the INT / removed 
immediately after the beginning of the cycle extraction KOP'a 
(Third front (dx) signal carries the M1 line / signal INT / in 
the log. 1), and the query interrupt is removed. The result is 
that "grab" twice in a row INT / CPU can not - that is, 
duration INT'a optimally. (I want to draw note that in this 
case it is arbitrary - and it makes no sense to consider the 
zeros in the duration INT'a, - he SYNCHRONOUS CPU). Activist, 
who published a schematic of a ZX-Revue, will be like a crow, 
which saw a shiny object, and seizing its beak (not your 
head!), took him in nest. Just because he's brilliant

(Regardless of what the subject,
- A wedding ring on the sill, the fuse from a grenade or simply 
color Spangle). What is he has found what it is

have and what wonderful, he did not understand.

   I would like to discuss INT / Pentagon'a.
If you forget about the phase INT'a (as it already
discussed), the INT - is the source (kalibrovannyy!) time. And 
this his hypostasis found not only in its period (Which is 
trivial), but also in duration. In SYNCHRONOUS circuitry is only

One source of time - it's quartz, - and
It generates the clock sequence. RC-chain - it's trivial for a 
time source. And, it should be added very bad. Explain why. 
Duration of time interval, which is formed monostable 
multivibrators on the RC-circuit is determined the following 
equation: 

   T (sec) = k * R * C = k * t; (t = R * C) In this relation 
(function, formula), three (TRI!) The source of the error: k - 
depends on the supply voltage (more than proportionately); C - 
depends on temperature and time (ie, aging); R - depends on the 
time (also older). And, with all Thus, the sampling error of + 
/ -1 tact. Once, years that way five years ago, in

repair (through the "Glitter") and received
Pentagon'y. They were strange knobs (on the variable resistor) 
INT'a. At first it was difficult to understand why it is done. 
Then he got it. Users so "caught" demos on their companies. As 
in radio - shook - and tuned

the desired station. And it is in the computer! The bottom line 
is that there is not and PENTAGON'e Can be the standard INT'a - 
ie, ALL Pentagon'y mutually curves. And if they

a time mutually lines, then
is nothing more than short-term success -
INT-it goes with the times! How can
leave a synchronous (on the RC-circuit) INT? A
very simple. Fixed resistors, the
Indeed, is not constant on the length of time. In "Mass 
radiobiblioteke" or in "Radio," You Of course, this does not 
subtract. Here we need Reference state standards. Even under 
conditions STORAGE (ie in the absence of

electrical / mechanical / thermal
loads) care face value of 15
years to 10%, and for some resistors, and
30%! Not to mention vysheupominavsheysya
Depending on the temperature (up to 50%
Some types of capacitors in the WORK
temperature range) and voltage (about 5% - tolerance of 
Nutrition for TTL-series). As a result, users

Pentagon'ov have foaming at the mouth to prove that their car 
right, and all others - the curves - and it is only among users 
Pentagon'a! You need this? A What is 10% for INT'a? This means 
that where he was 9mks, he became 10mks. AND

software compatibility demam flew down the drain. Analogue FINE 
- a variable resistor front! Sun rose, warmed the computer - it 
works so well in the evening, when it is dark - that way. 
Wagged resistor - and all the way through! 


   To summarize. Hardly pentagon'ovsky "... normal sync ..."
can be called normal. Because as those standards do not exist. 
But in KAY-e, he made just normal. Ie

All of the machines, as twins are identical. Another thing is 
that INT is formed differently than in Pentagon'e. 

   It is useful to think on a topic such as
TURBO-mode. When the scheme of formation,
Synchronously in CPU cycles, ie as
in KAY'e on 555IR16 - programs are not
Distinguish between the duration of INT'a
between modes TURBO / NORMAL! Ie programmer, when processing 
INT'a, there is no need think about how his program will work

in TURBO-mode - it will work just as well as in normal mode. 
This very important for the TURBO-machines.


   "In this regard, I have a rhetorical
question ... "- Is necessary to repeat
and propagate nonsense, just for the
because it is COMMON?

4. Description of work from the top of memory. " A
why do you this description? Exercise in
programming? Is that ... There is a very
much doubt that it is possible the traditional
Spectrum'ovskimi methods (technologies,
concepts, approaches) to effectively utilize 1Mbayt RAM. Ie 
probability obtaining a commercial software (ie, that

software, which makes the machine useful for
user) tends to zero. Widely
common argument (in the style of Parkinson's) that the software 
tends to occupy all available memory in Spectrum'e not quite 
true. If you are already commercial programs under 128Kb th car 
created for years, then how long it takes at harnessing 
1Mbayta? We are not talking about how to take all the memory, 
"something" - it is not working. Speech comes to effective use.


   Therefore, in KAY'e and 1MB is not available
"Naked" - as a mother gave birth - and there
given the structure of the memory of RAM - a,
Specifically, kvazidiskovod "C". If you go
fit, it will fly under the cat's tail. AND
yet the fact that the effectiveness of the use of "directly" 
1Mbayta be more (And useful for users) than kvazidiskovod.


   Apparently, the efficient use of
such amounts of RAM can only be under the operating system. And 
if you work with OS'yu, information about the real Physical 
devices (architecture) com Is unnecessary, because treatment 
are at OS'i actions, ie abstracted from

Real hard'a. It is widely known
IBM-ovsky approach. If you have an alternative (at a fairly 
deep conceptual level), it is not tomite - to share - very 
interesting. Me such approaches are unknown. I think those who 
are able to create something really worthwhile, commercial, 
useful for all, yet in a position to look into the scheme. Or 
in passport machines, where Table 1 shows

two bits A17 and A18 (capital letters).
For information: in kvazidiskovode seniority
bits of the opposite, ie senior is
A17.

5. "According to the shelves." Need to make a general 
observation that, indeed, you mentioned "white spots" in the 
description machines exist. In the "User Guide", which is 
currently being prepared for publication, these issues will be 
covered. Now the signals. 


   Q8. Signal, a freelance for the interface
Centronics. In other words, all machines
along this line comes the earth (GND). This
convenient from a technological point of view.
Easier to unsolder / Breeding flat interface cable with the 
pair alternating Signal data / Earth. " Therefore, if it

engage in hard'e, then the RS-semiconductors and
omig'ah will not work. (That would be
impossible "software" to get to this signal). Otherwise, there 
is nothing special. It's OK (open collector).

Active state, - "is connected to ground, ie, log. 0. To install 
the log. 0 must set bit * D1 port # 1FFD

in the log. 1. To reset this signal is
in a state of high impedance, and to
a log. 1, an external device
should be a resistor to +5 V. Ie normal
specifics of the scheme on "OK".

   AUTO. In fact, this INIT CENTRONICS'a. Just in the 
designation of signal names connector XS1 they were confused. 
Full-time signal centronics'a. Printer it is initialized 
(taking the original, specified in the passport, state). When 
you connect other devices can be used similar signals Q8, -

bit * D6 port # 7FFD.

   A18, A17 - already discussed. (Again
also, see Table. A passport machines).

   ROMS - signal inversion, "Rom.
When you install the log. 1 bit * D3 port
# 1FFD shadow region is changing places with
Basic 128 and TRDOS varies with BASIC 48.
(Or vice versa, depending on the jumper (jumper) JP10, 11,12).

   C1 - a signal that is not available to the programmer. 
Hardly a subject worthy of discussion. By this signal is 
blocked joystick port (Apparatno!) # 1F when accessing the 
music processor for reading (address # FFFD). Otherwise they 
would have clashed on the bus. 

5.1 "Management of upper memory." I doubt that even if this
will be included in the documentation, programmers will be able 
to anything to support the KAY-1024. Once again, that without 
OS'i this, in my opinion, impossible. And when

is OS' - is unnecessary.

5.2 "Connector RGB". Built-in RGB, in general, the explanations 
are not needed. It is made for greater flexibility in 
connecting board to the display device (usually TV). Based on 
past experience, most boards "picking"

in this place. Therefore undertaken
attempt somehow to get away from the soldering iron in the 
process of connecting TV-receiver. Those in the able to take 
advantage of this connector, the explanations are not needed. 
And those who Need clarification - it is better not to go there.

Ie a technological connector for professionals. On the 
architecture (ie the programming model) car, he did not 
influences. 

   In conclusion I would like to express gratitude for the 
substantive issues. I think that the explanations will be 
useful, - and not only to you.



                                  (C) Nemo



               APPENDIX


   Hello, Dear Editors,
ZXFormat. Wrote to you from the Novocherkassk
Hacker <Grey> (Gray), although at present
I have more time Musicmaker and coder, rather than
HaCKer. I am a member of the largest and
known in the city of Rostov-on-Don group
Omega Hackers Group. Perhaps someone
met my works: Omega Commander,
demo version of the game "The Way to Earth" or my
old hacker working for the signature "Bologov Gregory from * 
OHG *". I will not for long grovel, just say that I really

like the ZX-Format, and not only externally
just would like to see the section "Hardware"
more consistent with its name -
had more schemes.

   I hope my letter will shed some light on an old problem (or 
not problem) Pentagon-ovsky interrupts. More

when I had my ATM turbo and Spark, I
resigned to the fact that the Russian Spectrum
Pentagon is 128. Now I have a Pentagon
3 +, I have set myself another double
switch. One side of the switch interrupts branded on 
Pentagon-to-native LN1, and other closure 15 foot counter DN6 
(scheme Pent 3 +) on land, which is similar to 2 and 3/D11 on 
Pentagone 128, thereby getting full Pentagon interruption.


   "What's the matter?" - You ask. But the fact that the Firm 
Speccy Frame Scan is 312 lines, and the Pentagon 128 reason 
320. Therefore, the firm speccy shows 69,888 cycles per 
interruption, and Pent - 71,680 (according to the test in the 
JC 4.01). But then manifests itself one mistake

(Or otherwise) in some variants
scheme Pentagon 128. This input capacitor C6 to interrupt. If 
it is equal to 22N, then the test will have 71,649 cycles. 
(Ideally, the C6 has to be 6n8). Like anything terrible and 
there is a loss of 31 cycles and that the duration of the INT-a 
was equal to two (for test test 128). But in this case, provide 
a 100%-D re- capture interrupt (even without TURBO). But

this is not the worst.

   The fact that many coders are
advantage muLticoLor-ax and
curb the effects of auto-tuning, determining the Pentagon or 
not. Given this, we must make allowances + / - 60 cycles (there 
are cars, which C6 = 47n, which does not prevent multicoloRit, 
but floods test). Many coders take their cars for an ideal that 
is not a fact (because the majority of Pentagon 128 C6 = 22N).


   It suffers from a demonstration ENLIGHT 1996
- PRSG 128, refusing to multicolor
regular Pentagon interrupts
However, if the installation included TURBO, and then shut down 
- muLticoLoRiT, how cute. 

   Just want to say that the Pentagon
128 can not get branded interruption. Even if the decay of the 
form INT human clock, gathering scheme in

capacitors and resistors, will have to
hanging tree logic on 2 and 3 findings D11,
narrowing the scan up to 312 lines, or effects, made under 
psevdofirmennye interrupts will not work either on a normal 
Pent-e, nor on the normal corporate interrupts.



   In conclusion, I want to summarize all
above:

   1. On virtually any machine can
closing the counter reset line to ground
Pentagon to get an interrupt.

   2. C6 on the scheme Pentagon 128 should
be 6n8, not 22N, on the Pentagon-e must
be 71 680 cycles 320 * 224 = 71680, not
71649. On the company's Speccy - 69,888 cycles
312 * 224 = 69888; 224 - number of ticks in
scan line.

   3. When auto-tuning by interrupting
consider the error of + / -60 cycles or
even more (because of the different values of the capacitor 
C6). 

   4. muLticoLor-s under the brand interrupts can not be done 
at the Pentagon 128 psevdofirmennymi interrupts.



   I hope my letter will help to understand this question for 
those who just this still does not know.


   In my method, brought to mind the three left (in terms of 
Pentagon 128) machine. They are now in a position of the switch 
work perfectly INSULT, PSG wins, SATISFACTION, while the other 
is absolutely not mow SHOCK, ECHOLOGY (naturally, except border 
under separate field memory SHOCKe)



                   14.12.06 Grey OF * OHG *

P.S. Huge request, who knows how
get rid of the lines in TURBO mode on
scheme of V.M.G. or ON-LINE N2, tell us about it in the pages 
of ZF or write to: 346 421 Rostov region. Mr. Novocherkassk, 
ul. Budennovsk 237 / 2 square. 21 Bologova GA


P.P.S. Thank you in advance Received ZX-Format-a, and I hope 
that my letter be published in forthcoming issues ZF. 



HELP

Subject PENTAGON'ov also discussed in
letters N 34 and 43.